T-NGA paper accepted to ICASSP
by Tobi
We are happy to announce that our paper S. Wang, Y. Hu, and S-C. Liu, “T-NGA: Temporal network grafting algorithm for learning to process spiking audio sensor events” (see preprint here) has been accepted to the 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (2022 ICASSP), the top audio signal processing conference worldwide.
This paper shows that the Neural Grafting Architecture developed originally by Yuhuang Hu can be used to train a speech vs. noise classifier purely with pairs of unlabeled audio recordings from a Dynamic Audio Sensor (DAS) silicon cochlea to achieve accuracy within 5% of the conventional DNN driven by Nyquist-sampled audio signals. This paper opens a new pathway to efficiently train new DAS audio applications.
Congratulations to Phd students Shu Wang and Yuhuang Hu.